Abstract: The multi-site monitoring,high data analysis workload and manual fault recognition manner makes the wind turbines operational reports months delay.An automatic identification algorithm for wind turbine gearbox based on amplitude modulation is proposed in this paper.According to the unsteady feature of the wind turbine speed,the time-frequency analysis is carried out on the gearbox vibration acceleration signal firstly to obtain the instantaneous speed of the rotating,then the resampling of order tracking analysis is conducted to transform the equal time signal sequences to equal angular ones,and then the larger coefficient value is selected between the meshing and the double meshing frequency in the frequency domain.The multi frequency coefficient value spacing by rotating speed is summed and the ratio divided by larger meshing frequency is deemed as the feature value to character the state of the gearbox failure.The effective of distinguish the faults and normal state is shown by gearbox vibration data analysis results under both constant and variable speed condition,and the stability is also shown based on different speed simultaneously.